Cranium：神经网络库-C语言

#include "../src/std_includes.h" #include "../src/matrix.h" #include "../src/function.h" #include "../src/layer.h" #include "../src/network.h" #include "../src/optimizer.h" int main(){ // test backpropagation correctness srand(time(NULL)); size_t hiddenSize[] = {2}; void (*hiddenActivations[])(Matrix*) = {sigmoid}; Network* network = createNetwork(2, 1, hiddenSize, hiddenActivations, 2, softmax); float** data = (float**)malloc(sizeof(float*) * 1); data[0] = (float*)malloc(sizeof(float*) * 2); data[0][0] = 1; data[0][1] = 0; DataSet* dataR = createDataSet(1, 2, data); network->connections[0]->weights->data[0] = .5; network->connections[0]->weights->data[1] = 1; network->connections[0]->weights->data[2] = -1.5; network->connections[0]->weights->data[3] = .25; network->connections[1]->weights->data[0] = .25; network->connections[1]->weights->data[1] = 1; network->connections[1]->weights->data[2] = -.5; network->connections[1]->weights->data[3] = -1; forwardPassDataSet(network, dataR); assert(network->layers[2]->input->data[0] >= .474 && network->layers[2]->input->data[0] <= .475); assert(network->layers[2]->input->data[1] >= .520 && network->layers[2]->input->data[1] <= .530); float** classes = (float**)malloc(sizeof(float*) * 1); classes[0] = (float*)malloc(sizeof(float*) * 2); classes[0][0] = 0; classes[0][1] = 1; DataSet* classR = createDataSet(1, 2, classes); batchGradientDescent(network, dataR, classR, CROSS_ENTROPY_LOSS, dataR->rows, 1, 0, 0, 0, 1, 0, 1); assert(network->connections[1]->weights->data[0] <= -.04 && network->connections[1]->weights->data[0] >= -.05); assert(network->connections[1]->weights->data[1] >= 1.29 && network->connections[1]->weights->data[1] <= 1.30); assert(network->connections[1]->weights->data[2] <= -.84 && network->connections[1]->weights->data[2] >= -.85); assert(network->connections[1]->weights->data[3] <= -.65 && network->connections[1]->weights->data[3] >= -.66); assert(network->connections[0]->weights->data[0] >= .58 && network->connections[0]->weights->data[0] <= .59); assert(network->connections[0]->weights->data[1] >= .94 && network->connections[0]->weights->data[1] <= .96); assert(network->connections[0]->weights->data[2] <= -1.4 && network->connections[0]->weights->data[2] >= -1.6); assert(network->connections[0]->weights->data[3] >= .24 && network->connections[0]->weights->data[3] <= .25); destroyDataSet(dataR); destroyDataSet(classR); destroyNetwork(network); // test on XOR ([off on] ordering) int i; data = (float**)malloc(sizeof(float*) * 4); for (i = 0; i < 4; i++){ data[i] = (float*)malloc(sizeof(float) * 2); } data[0][0] = 0; data[0][1] = 0; data[1][0] = 0; data[1][1] = 1; data[2][0] = 1; data[2][1] = 0; data[3][0] = 1; data[3][1] = 1; DataSet* trainingData = createDataSet(4, 2, data); classes = (float**)malloc(sizeof(float*) * 4); for (i = 0; i < 4; i++){ classes[i] = (float*)malloc(sizeof(float) * 2); classes[i][0] = (data[i][0] == 0 && data[i][1] == 0) || (data[i][0] == 1 && data[i][1] == 1) ? 1 : 0; classes[i][1] = classes[i][0] == 1 ? 0 : 1; } DataSet* trainingClasses = createDataSet(4, 2, classes); size_t hiddenSize2[] = {3}; void (*hiddenActivations2[])(Matrix*) = {tanH}; Network* network2 = createNetwork(2, 1, hiddenSize2, hiddenActivations2, 2, softmax); printf("\nTESTING ON XOR:\n"); printf("Starting accuracy of %f\n", accuracy(network2, trainingData, trainingClasses)); batchGradientDescent(network2, trainingData, trainingClasses, CROSS_ENTROPY_LOSS, trainingData->rows, .3, 0, .01, .9, 1000, 1, 1); printf("Final accuracy of %f\n", accuracy(network2, trainingData, trainingClasses)); // test on above or below y=x^2 [x y ordering] [below above ordering] float** dataF = (float**)malloc(sizeof(float*) * 100); for (i = 0; i < 100; i++){ dataF[i] = (float*)malloc(sizeof(float) * 2); dataF[i][0] = 1.0 * (i + 1) * (rand() % 50); dataF[i][1] = 1.0 * (i + 1) * (rand() % 50); } DataSet* trainingDataF = createDataSet(100, 2, dataF); float** classesF = (float**)malloc(sizeof(float*) * 100); for (i = 0; i < 100; i++){ classesF[i] = (float*)malloc(sizeof(float) * 2); classesF[i][0] = dataF[i][0] * dataF[i][0] <= dataF[i][1] ? 1 : 0; classesF[i][1] = classesF[i][0] == 1 ? 0 : 1; } DataSet* trainingClassesF = createDataSet(100, 2, classesF); size_t hiddenSizeF[] = {3}; void (*hiddenActivationsF[])(Matrix*) = {tanH}; Network* networkF = createNetwork(2, 1, hiddenSizeF, hiddenActivationsF, 2, softmax); printf("\nTESTING ON PARABOLA:\n"); printf("Starting accuracy of %f\n", accuracy(networkF, trainingDataF, trainingClassesF)); batchGradientDescent(networkF, trainingDataF, trainingClassesF, CROSS_ENTROPY_LOSS, 20, .01, 0, .01, .5, 1000, 1, 1); printf("Final accuracy of %f\n", accuracy(networkF, trainingDataF, trainingClassesF)); // test on regression on y=x^2 + 15 float** dataReg = (float**)malloc(sizeof(float*) * 1000); for (i = 0; i < 1000; i++){ dataReg[i] = (float*)malloc(sizeof(float) * 1); dataReg[i][0] = rand() % 20; } DataSet* trainingDataReg = createDataSet(1000, 1, dataReg); float** classesReg = (float**)malloc(sizeof(float*) * 1000); for (i = 0; i < 1000; i++){ classesReg[i] = (float*)malloc(sizeof(float) * 1); classesReg[i][0] = dataReg[i][0] * dataReg[i][0] + 15; } DataSet* trainingClassesReg = createDataSet(1000, 1, classesReg); size_t hiddenSizeReg[] = {20}; void (*hiddenActivationsReg[])(Matrix*) = {relu}; Network* networkReg = createNetwork(1, 1, hiddenSizeReg, hiddenActivationsReg, 1, linear); printf("\nTESTING ON PARABOLA REGRESSION:\n"); float** oneEx = (float**)malloc(sizeof(float*)); oneEx[0] = (float*)malloc(sizeof(float)); oneEx[0][0] = 5.0; DataSet* oneExData = createDataSet(1, 1, oneEx); forwardPassDataSet(networkReg, oneExData); printf("Initially maps 5 to %f but should map to 40\n", networkReg->layers[networkReg->numLayers - 1]->input->data[0]); batchGradientDescent(networkReg, trainingDataReg, trainingClassesReg, MEAN_SQUARED_ERROR, 20, .01, 0, .001, .9, 500, 1, 1); forwardPassDataSet(networkReg, oneExData); printf("After training maps 5 to %f but should map to 40\n", networkReg->layers[networkReg->numLayers - 1]->input->data[0]); // test parameter set ParameterSet params = {networkReg, trainingDataReg, trainingClassesReg, MEAN_SQUARED_ERROR, 20, .01, 0, .001, .9, 200, 1, 1}; printf("\n"); optimize(params); destroyDataSet(trainingData); destroyDataSet(trainingClasses); destroyNetwork(network2); destroyDataSet(trainingDataF); destroyDataSet(trainingClassesF); destroyNetwork(networkF); destroyDataSet(oneExData); destroyDataSet(trainingDataReg); destroyDataSet(trainingClassesReg); destroyNetwork(networkReg); return 0; }